The New Sesquipedalian

Elsi Kaiser (University of Southern California) will give a colloquium today (Friday Nov. 1) at 3:30 PM in the Greenberg Room, followed by a departmental social.

PERSPECTIVE-TAKING EFFECTS ON REFERENCE RESOLUTION: PRONOUN INTERPRETATION AND (NON-)SPEAKER-ORIENTATION

Abstract: Tracking different perspectives is an important but controversial aspect of language comprehension. I will present a series of experiments on free indirect discourse (FID), which hinges on the ability to recognize perspective shift. Free indirect discourse presents characters’ speech/thoughts without quotes or embedded clauses (ex.1b,2b). How do readers realize that the perspective has shifted from the narrator to one of the characters? Narratologists have identified various cues to FID, including expressives (e.g. poor girl, that idiot) and adverbials of possibility (e.g. possibly, probably) (Fludernik 1993, Bortolussi and Dixon 2003, i.a.). This work connects in important ways to linguistic research on (non-)speaker-oriented readings of expressives and appositives (e.g. Potts 2005, Harris & Potts 2010, Harris 2012; Gutzmann 2013, see also Sharvit 2008 on FID).

(1a) Mary looked woefully at Kate. She was sick. [no cue]
(1b) Mary looked woefully at Kate. Poor girl; she was sick. [FID cue]
(2a) Luke glanced at Tom warily. He’d put toothpaste in the shampoo bottle again. [no cue]
(2a) Luke glanced at Tom warily. He’d probably put toothpaste in the shampoo bottle again. [FID cue]

I present a series of experiments (including visual-world eye-tracking) investigating comprehenders’ sensitivity to different kinds of FID cues, in order to test whether (and how successfully) different cues trigger perspective shifts and how these shifts interact with pronoun interpretation. We used sentences similar to ex.(1-2), designed so that processing the FID cue affects pronoun interpretation. In plain sentences (1a,2a), the pronoun in the second sentence is ambiguous, but in sentences with FID triggers (1b,2b), if people recognize the perspective-shift, they should interpret the pronouns as object-referring. In addition to probing participants’ pronoun interpretations, we also asked them to indicate whose point-of-view is represented in the second sentence.

As a whole, the results show that people can successfully recognize subtle linguistic cues as signaling a shift from the narrator to a character’s perspective, and that these cues have rapid effects on real-time comprehension. However, not all cue types are equally effective, which raises interesting questions regarding the perspective-shifting ability of epithet/expressive-type cues vs. epistemic cues like ‘probably.’ The strongest cues elicit high rates of ‘character’s perspective’ interpretations, around 70-80%, which suggests that perspective-shifting may not be as costly as is often assumed. Furthermore, our findings pose challenges for syntax-centered theories of pronoun resolution (e.g. the well-known ’subject bias’) and have implications for our understanding of referent salience/accessibility.

Time permitting, I will also discuss data from Finnish, a language with a richer anaphoric paradigm than English. The pronominal paradigms in colloquial and formal/written Finnish are different, and I will discuss how shifting between these two registers (as signaled by choice of anaphoric form), can signal a perspective shift (FID).

All are welcome!

Upcoming colloquia:
November 8: Roger Levy
November 15: David Beaver

Please join the Phonology Workshop today (Friday November 1) at noon in the Greenberg Room for a talk by Brianna Kaufman (UC Santa Cruz).

Turkish reduplication, fixed segmentism, and DM/OT

This talk will address the pattern of emphatic reduplication in Turkish e.g. kara “black” to kapkara “very black.” The picture is complicated by the fact that the fixed segment that intervenes between the reduplicant and the base alternates between four segments. Which of the four fixed segments surfaces is conditioned in part by co-occurrence restrictions. The standard approaches to reduplication within DM are unable to derive the correct outputs because they do not reference the right level of prosody, and because they fail to account for attested phonological alternations and variations. The proposal at hand will expand a blended model of DM and OT (á la Haugen 2008; 2011) in order to account for the very complicated case of fixed segmentism in Turkish.

Free lunch to follow!

Stephanie Shih will be giving the oral presentation of her dissertation this Tuesday, Nov. 5, at 9AM in the Greenberg Room. (The format for this open part of the oral exam is a 30-45 minute talk by the PhD candidate followed by questions from those attending, for a total of no more than one hour. Please arrive promptly!)

Towards Optimal Rhythm

This thesis argues that rhythmic well-formedness preferences contribute to conditioning extra-phonological operations, providing evidence from patterns in language use that constraints on phonological constructs are at work in the assessment of competing morphosyntactic variants. The results of the thesis call into question a fundamental assumption in standard models of grammar and of language production, that the (morpho)-syntax has no forward knowledge of low-level metrical or segmental phonology (e.g., “Phonology-free Syntax”). Such a restriction between syntax and phonology arises from two sources in linguistic models: modularity between components of grammar and unidirectionality in information flow between these grammatical modules. The view taken in this thesis maintains that, counter to previous assumptions, information flow is not strictly unidirectional in that constraints operating over phonological information apply in the morphosyntactic component of grammar. Instead, it is the modularity of grammatical components that gives rise to the depth-of-information-access limitations that remain in the interaction between phonology, phonetics, and morphosyntax.

Evidence against unidirectionality is presented from two empirical case studies of rhythm’s role in word choice (i.e., personal name choice) and syntactic construction choice (i.e., genitive alternation) in English. The case studies demonstrate that rhythmic optimization, in addition to other phonological well-formedness preferences such as anti-homophony, are active in word and construction variation in syntagmatic contexts. It is furthermore shown that the effect of rhythm is closely tied to semantic factors such as animacy, which reveals that rhythm must interact with non-phonological constraints in the system.

Allowing bidirectionality in a model of the phonology-syntax interface allows for more interaction, but the relationship between each component of language is still crucially constrained by modularity of the grammatical components. It is argued that in the assessment of morphosyntactic competitors, phonological and phonetic competitors—that is, the potential outputs of phonology and phonetics—are not available. Rhythm offers a natural test case of the availability of underlying versus post-lexically-specified information via the distinction in stress properties of lexical (content) and grammatical (function) words. A large-scale corpus study of content and function word stress in conversational American English is presented. Results of the study point to complex differences between word categories in terms of underlying and surface stress properties. These differences in stress not only trigger differences in rhythmic optimization by word category but they also demonstrate that forward knowledge of surface-assigned phonetic stress is unavailable at the point of rhythmic optimization between morphosyntactic constructions. In contrast, evidence from end weight phenomena suggests that lexically-encoded information about underlying phonological stress is available during morphosyntactic computation.

The view that emerges from the empirical studies in this thesis is one that weakens the claim of Phonology-free Syntax. Phonological well-formedness conditions apply at all levels of grammar, including the syntactic component. Restrictions on phonology-syntax interaction arise from modularity in the system, meaning that morphosyntactic choices are made without competition with phonological or phonetic output candidates. Positing the maintenance of modularity in the grammar explains the relative rarity of phonological influences on extra-phonological operations. Phonological constraints that are most active will necessarily be ones that regulate syntagmatic effects that occur when words combine in the morphosyntactic module, and these phonological constraints—including the propensity towards optimal rhythm—must compete for satisfaction against the other active, non-phonological pressures.

Oral exam committee: Arto Anttila and Sharon Inkelas (Co-advisors), Joan Bresnan, Dan Jurafsky
University oral exam chair: Herb Clark (Psychology)

At the next SMircle meeting (4PM, Tuesday November 5, Greenberg room), Maziar Toosarvandani from UCSC will talk about agreement in Zazaki. His title and abstract are below.

Agreement in Zazaki and the nature of nominal concord
Maziar Toosarvandani (UCSC), joint work with Coppe van Urk (MIT)

At first glance, agreement in the verbal domain looks strikingly different from agreement in the nominal domain in Zazaki, a Northwestern Iranian language of Turkey. Nonetheless, we argue that these two agreement systems are established by a single syntactic operation, which we identify as Agree (Chomsky 2000). We do this by showing that both verbal agreement and nominal concord are subject to the same restriction: phi-agreement with a DP bearing oblique case is blocked, a constraint familiar from other languages (Rezac 2008, Preminger 2011). Thus, Zazaki provides an argument that agreement is derived by the same mechanism regardless of syntactic domain (Mallen 1997, Carstens 2000, Baker 2008). We also propose that the Agree operation that derives both nominal concord and verbal agreement in Zazaki is bidirectional in nature (Adger 2003, Baker 2008), with a preference for operating downward first, before it operates upward. In the verbal domain, this property has been shown to arise from Agree applying as soon as it can within a cyclic domain (Béjar and Rezac 2009). We take Zazaki to provide confirmation from the nominal domain that Agree operates in this cyclic fashion.

Please join us for a colloquium given by Roger Levy (UCSD) next Friday, November 8, at 3:30PM in the Greenberg Room (followed as always by a social).

The ecology of the binomial construction: processing, pragmatics, and efficiency

Abstract: The binomial construction, X Conj Y (e.g., “salt and pepper”, “hit and run”) is ideally suited to investigating two central issues in the contemporary study of language: (1) the relationship between abstract, rule-based and direct, exemplar-based knowledge, and (2) what factors govern speaker choice.  In this talk I combine corpus analysis, controlled experimentation, and computational modeling to shed light on these two issues.  The first part of the talk explores the contributions of different knowledge sources in preferences regarding binomial ordering.  It is well-established that several types of abstract knowledge — semantic, lexical, and phonological — are jointly predictive of ordering preferences for binomials found in corpora.  However, it is unclear how this abstract knowledge is represented and used in the mind of the speaker.  We address this problem by asking (a) whether speakers’ real-time processing of novel binomials they have never before encountered reflects the application of these abstract information sources, and (b) whether, in real-time processing of attested binomials, direct experience trumps this abstract knowledge.  The answer to both these questions is “yes”, supporting models of linguistic knowledge that can both store and reuse exemplars and abstract out generalizations from those exemplars.

In the second part of the talk, I investigate a challenge to pragmatic theory posed by one class of binomials, exemplified by naturally occurring sentences such as “We sell roses and flowers for Mother’s Day” or “Tattoi quickly slipped into the meat and beef store”.  In semantic theory, “roses and flowers” should have the same literal meaning as just “roses”, but its actual interpretation is indiscernible from that of “roses and other flowers”.  Alternations such as “roses and flowers”/”roses and other flowers” turn out to be widespread and historically stable.  This yields the following challenge: in a system of alternative utterances including “roses” and “roses and other flowers”, what properties of a natural language interpretation system are required to robustly resolve the pragmatically enriched meaning of “roses and flowers” to be equivalent to the latter, rather than the former, alternative?  Meeting this challenge turns out to require an extension of state-of-the-art theories of rational speech acts, and sheds light into the typology of generalized conversational implicatures.

Rob Podesva presented “The Other California English: Sociolinguistic Variation in the Central Valley” in the Lectures in Language and Linguistics series at California State University, Bakersfield on October 30.

Chris Potts presented “Conversational Implicature: Interacting with Grammar” as a Linguistics Club Lecture at the University of Illinois, Urbana-Champaign on October 28.

Stanford Alum Alessandro Jaker (Alaska Native Language Center, Stanford PhD 2012) will present “Selection and Blocking in the Northeast Dene Verb: Some Counterexamples to Level Ordering” at the 2nd American International Morphology Meeting at UCSD on November 8-10.

At Phonology 2013 in Amherst on November 9-10 the following Stanford linguists will speak:

An angel came down for a meeting of the American Philosophical Association. Greeting the assembled philosophers, the angel offered to answer a single question for them. Immediately the philosophers set to arguing about what they should ask. So the angel said, “Alright, you figure out what you want to ask. I’ll come back tomorrow.” And he left the philosophers to deliberate.

Some of the philosophers favored asking conjunctive questions, but others argued persuasively that the angel probably wouldn’t count this as a single question. One philosopher wanted to ask “What is the best question to ask?”, in the hope that some day another angel might make a similar offer, at which point they could then ask the best question. But this suggestion was rejected by those who feared that no such opportunity would arise and did not want to waste their only question.

Finally, the philosophers agreed on the following question: “What is the ordered pair whose first member is the best question to ask, and whose second member is the answer to that question?” Satisfied with their decision, the philosophers awaited the angel’s return the next day, whereupon they posed their question. And the angel replied: “It is the ordered pair whose first member is the question you just asked, and whose second member is the answer I am now giving.” And then he disappeared.

[From a list compiled by Chrissy Stockton – thanks to Tania Rojas-Esponda for finding it.]